XAID: Explainable Artificial Intelligence for Alzheimer’s Disease Diagnosis and Progression Quantification from Multi-Modality Data

Population aging is one of the increasingly serious social problems all over the world. Ensuring the aging population has timely access to the high-quality healthcare is a desirable yet urgent goal for our whole community. Particularly, as an implicitly developing neurodegenerative disease, Alzheimer’s disease (AD) has become one of the most intractable problems in aging population. In clinical practice, AD is usually characterized by some degenerative dementia appearances, e.g., memory impairment, aphasia, agnosia, impairment of visuospatial skills, executive dysfunction, and changes in personality and behavior. Although AD has caused significant collateral damage on the aging population, the internal etiology of this disease remains largely unknown. With the recent development of artificial intelligence (AI) technologies, AI-based analytical models for disease analysis (e.g., patient-level disease classification, region-level morphological changes and pixel-level structure segmentation) have achieved huge advances. However, the AI models are often criticized as black boxes due to the lack of interpretation, which is of significance for gaining trust and ironing out concerns from medical professionals. In this project, we aim to develop state-of-the-art explainable AI technologies for AD diagnosis and progression quantification from the multimodality data, including heterogeneous feature extraction, quantitative anatomical structure analysis and interpretable biomarker discovery. These AI-driven models will be integrated into a unified clinical decision support system (CDSS) for facilitating AD’s diagnosis, quantification, biomarker interpretation, and treatment.